A-Share Smart Portfolio Monitor and Investment Research Assistant

The problem

A common pain point in A-share monitoring is not a lack of data, but data scattered across quotes, filings, funding flows, technical indicators, and sentiment. Portfolio screenshots are often unstructured, manual threshold checks create friction, and push alerts can become noisy during non-trading hours. This skill organizes a portfolio pool, alert rules, data-source calls, and output formats into a reusable agent workflow for engineering and research workflows.

How it works

• Portfolio input: uses the Read tool to extract codes, share counts, cost basis, and P&L from screenshots; manual entry supports real, simulated, and watchlist tags.
• Data integration: calls neodata-financial-search and westock-data for quotes, K-lines, MACD/KDJ, main and northbound flows, financials, and valuation.
• Anomaly monitoring: triggers on price, volume, turnover, funding, technical patterns, and announcements, with thresholds adjustable through natural language.
• Session gating: automatic alerts run only during trading-day intraday windows; lunch break, post-market, and non-trading days stay silent while manual queries remain available.
• Analysis output: combines technical, funding, fundamental, and news dimensions, then frames short-term, mid-term, and long-term references.
• Fallback handling: on timeouts or empty results, it retries once against fallback sources and marks missing dimensions instead of aborting the whole analysis.

Boundaries and cautions

It is better suited for information aggregation, rule-based monitoring, and research drafts, not investment advice. The portfolio pool is session-oriented; cross-session recovery still depends on screenshots, lists, or memory modules. Dragon-tiger list, northbound flow, and related coverage is mainly for Shanghai and Shenzhen A-shares; HK and US markets are supported but with lower depth. For latency-sensitive strategies, treat it as an assist layer rather than a direct trading signal source.